An artificial intelligence program created by researchers at Stanford University can predict if someone will die within three months to a year, thereby improving end-of-life care.
This artificial intelligence program can help improve palliative care.
Everyone dies—what’s uncertain is when and how we die. To some, this information is something that they’d rather not know. Others, however, might take some comfort and solace from the knowledge. Scientists have now found a way to give these people a chance to find out whether or not death is imminent, specifically in cases where palliative care is a necessity.
Knowledge like this may make some people feel uncomfortable, but others can use the information to tie up loose ends. This can also mean that the timing for end-of-life care for critically ill patients can be much better than it ever was. According to a new study, the AI program was able to accurately predict mortality in 90 percent of cases.
Machine learning enabled the AI to digest data from 200,000 patients.
Predicting a person’s mortality isn’t as simple as looking at someone and determining that they’re not looking too good. There are factors such as age, family health history, how well the patient responds to drugs, and the nature of the illness itself. Doctors also have their own biases and preferences when it comes to coming up with a prognosis, and their predictions can range from accurate to way off base.
The researchers mention in their study that while 80 percent of Americans would like to spend the last days of their lives at home, only 20 percent get their wish. This is because prognoses aren’t accurate, and patients receive more treatment at medical facilities instead of living out their last days at home. Ken Jung, one of the researchers, wants palliative care to be able to ensure that the patient’s wishes are made known and fulfilled.
The researchers fed data from juvenile and adult patients’ electronic health records (EHR) into the AI program to facilitate deep learning. Among two million patients, the researchers were able to find 200,000 patients who fit into their study. The program then studied the records of 160,000 of these patients, working under the direction to “predict the mortality of that patient within 12 months from that date, using EHR data of that patient from the prior year.”
Families may be more equipped to get affairs in order before the condition of their loved ones worsens.
The program then assessed the remaining 40,000 patients. It was able to successfully predict patient mortality within three months to a year in 90 percent of cases. Meanwhile, 95 percent of patients that were found to be unlikely to die within the three- to 12-month timeline did indeed live beyond a year.
“The idea behind using the algorithm is...so palliative care specialists can reach out to patients the algorithm recommends evaluation for. Then, the team would review the patient’s medical history and contact their physician about whether assistance is needed.
“Only after [receiving a] green light would the palliative care doctors contact the patients,” Jung told Newsweek. Thus, by predicting mortality, patients and their family members can discuss important matters before the patients’ condition worsens and they end up unable to make decisions on things that mattered to them.
Get weekly science updates in your inbox!